Add 3D+ input support for fp8 rowwise GEMM #2845
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
|
This pull request was exported from Phabricator. Differential Revision: D59671644 |
jianyuh
added a commit
to jianyuh/FBGEMM
that referenced
this pull request
Jul 15, 2024
Summary: Pull Request resolved: pytorch#2845 The input activation can be 3D+, with first dimension as batch dimension. - 1. If the input tensor is {M, K}, the output tensor is {M, N}. - 2. If the input tensor is {b, M, K}, the output tensor is {b, M, N}. This Diff adds the support, to match nn.Linear function / matmul supports. Differential Revision: D59671644
jianyuh
force-pushed
the
export-D59671644
branch
from
e626670
to
ea5b36e
Compare
|
This pull request was exported from Phabricator. Differential Revision: D59671644 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D59671644 |
jianyuh
added a commit
to jianyuh/FBGEMM
that referenced
this pull request
Jul 15, 2024
Summary: Pull Request resolved: pytorch#2845 The input activation can be 3D+, with first dimension as batch dimension. - 1. If the input tensor is {M, K}, the output tensor is {M, N}. - 2. If the input tensor is {b, M, K}, the output tensor is {b, M, N}. This Diff adds the support, to match nn.Linear function / matmul supports. Differential Revision: D59671644
jianyuh
force-pushed
the
export-D59671644
branch
from
ea5b36e
to
bdb30fc
Compare
|
This pull request was exported from Phabricator. Differential Revision: D59671644 |
jianyuh
added a commit
to jianyuh/FBGEMM
that referenced
this pull request
Jul 16, 2024
Summary: Pull Request resolved: pytorch#2845 The input activation can be 3D+, with first dimension as batch dimension. - 1. If the input tensor is {M, K}, the output tensor is {M, N}. - 2. If the input tensor is {b, M, K}, the output tensor is {b, M, N}. This Diff adds the support, to match nn.Linear function / matmul supports. Reviewed By: sijiac, jiawenliu64, jwfromm Differential Revision: D59671644
jianyuh
force-pushed
the
export-D59671644
branch
from
bdb30fc
to
73d6565
Compare
Summary: Pull Request resolved: pytorch#2845 The input activation can be 3D+, with first dimension as batch dimension. - 1. If the input tensor is {M, K}, the output tensor is {M, N}. - 2. If the input tensor is {b, M, K}, the output tensor is {b, M, N}. This Diff adds the support, to match nn.Linear function / matmul supports. Additionally, this Diff adds a FP8 architecture check: ``` functools.lru_cache def check_fp8_arch() -> None: arch_major = torch.cuda.get_device_properties(torch.cuda.current_device()).major arch_minor = torch.cuda.get_device_properties(torch.cuda.current_device()).minor if torch.version.cuda and arch_major < 9: raise Exception("FP8 can only work on Nvidia H100+ GPUs with sm90+ support!") if torch.version.hip and (arch_major < 9 or arch_minor < 4): raise Exception( "FP8 can only work on Nvidia MI300x+ GPUs with gfx942+ support!" ) ``` Reviewed By: sijiac, jiawenliu64, jwfromm Differential Revision: D59671644
|
This pull request was exported from Phabricator. Differential Revision: D59671644 |
jianyuh
force-pushed
the
export-D59671644
branch
from
73d6565
to
7f8ea83
Compare
|
This pull request has been merged in 903e928. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
The input activation can be 3D+, with first dimension as batch dimension.
This Diff adds the support, to match nn.Linear function / matmul supports.
Differential Revision: D59671644